Anonymity + metadata protection

[0chroma]

In the initial use case of tight-knit friend groups, knowing people's IPs should be ok. Eventually for more global, publicly accessible use cases, it will be required to have an option for anonymity at the cost of sync speed. People who are worried about metadata-centered data collection would also require this feature.

[0chroma]

Potential options to investigate

• Yggdrasil not anonymous
• Reticulum relies on i2p
• Noise Protocol see Identity Hiding
• Lokinet uses oxen, see note on Session below
• Gnunet may still too early in development to use, but promising - see secushare faq
• ipv8
• pinecone not anonymous
• Hosted relays funded by an opencollective of some kind like what syncthing does?

Looking at other apps

Many other apps seem to just route over Tor/i2p. It probably would be way too slow, especially when low latency is required ie games + group calls. You also wouldn't be able to accept connections either. It's just not well optimized for decentralized multicast use cases like InductionDB unfortunately. Briar took this route, but I'm not sure how well it works.

Session uses their oxen routing network for messages, but still uses direct connections for calls at time of writing. In general I don't like the idea of involving cryptocurrencies, I'd rather the network be self-sustaining such that every InductionDB instance is contributing to the network. More info on oxen/lokinet here

Need to read about how Matrix is doing it still

Some apps I want to research a bit closer:

• Reticulum MeshChat and Sideband - both work over Reticulum DHT likely doesn't scale
• SimpleX uses servers

both seem to have voice calling as a feature, I'm not sure if/how they protect IP exposure though.

Other places to investigate

There's a discussion on Freenet that is trying to figure out the same thing: freenet/freenet-core#813

Potential chart to investigate: https://bkil.gitlab.io/secuchart/

[0chroma]

I'm just gonna linkdump more stuff I need to read into:

Dat Protocol/Dat Rust
Aether
Holochain
RLN
Wesh/Berty

[0chroma]

I think for now I'm just gonna be stupid and continue with implementation, libp2p should (fingers crossed) allow us to abstract this problem so that regular TCP/UDP can be swapped out in the future

[0chroma]

Another option would be to implement onion-routing at the libp2p layer, providing an alternative to libp2p-kad-dht , example

or just provide a libp2p tor transport

[0chroma]

I've been thinking more about this, wondering if there's a way to encrypt the peer routing DHT and obfuscate CRUD operations to it through some sort of mixnet or other strategy. I think this is something I probably shouldn't focus on rn though, if I structure the code flexibly enough I can always try that later. It might also be better implemented in something like libp2p or a dedicated library on top of it. I think gnunet may do something like this but I'm not sure.

for an MVP it might be better to just offer relays like syncthing or signal do, or routing inductiondb traffic through tor/i2p and having routing handled by hidden services for nationstate requirements. There's also more experimental options like cjdns.